OKX Ventures Investment Monthly Report: The imagination of Crypto + AI is immense, having invested in multiple projects such as iO.NET and Myshell

Author: OKX Ventures

With the popularity of the internet and the acceleration of digitalization, massive amounts of data have become more accessible and usable. At the same time, the development of cloud computing and distributed computing technologies has made it easier and more cost-effective to process this data, providing richer resources for the training and optimization of AI models. This enables the development of smarter and more complex systems, driving the growth of the AI sector, especially in areas such as deep learning, natural language processing, and computer vision, where applications are widespread.

In the past year, the AI sector has exploded, led by ChatGPT and Nvidia. For instance, ChatGPT reached over 100 million monthly active users in just two months, becoming the fastest-growing application in internet history. Additionally, its recently launched Text To Video product, Sora, has garnered significant market attention, leading the AI video trend. Nvidia's net profit surged by 581% year-on-year in 2023, with its stock price skyrocketing from around $100 at the beginning of the year to over $800 now, increasing its market value by over a trillion dollars. Furthermore, OpenAI's founder is even looking to raise $7 trillion to create chips, among other new AI startup projects emerging continuously.

Against this backdrop, the new trend of AI is beginning to permeate the crypto industry. OKX Ventures believes there is a natural coupling between AI and the Crypto sector. In addressing decentralized computing power, data ownership, enhancing AI transparency, and model training inference, Crypto is an important complement to AI. Currently, Crypto + AI is becoming one of the most exciting frontier innovations.

Crypto + AI Applications Are Widespread

According to OKX Ventures, there are currently a large number of developers and researchers worldwide dedicated to the development and application of AI technology, forming a large and active community. The emergence and collaboration of these talents are promoting continuous innovation and progress in AI technology, especially as the Crypto + AI sector is becoming the most noteworthy Alpha track within this industry cycle.

The most direct use case of Crypto is to help various production factors in AI scenarios achieve assetization. For example, AI Agents, decentralized computing power, and data are all areas that can be assetized first and rapidly exploded.

Regarding AI Agents, OKX Ventures believes there will be three main innovative directions based on AI agents in the future. The first is the intelligent and agent-based management of assets, allowing assets to autonomously act and interact, thereby expanding their use cases and value. The second is interoperability between agents, establishing standards and frameworks that enable different AI agents to interact and collaborate efficiently, improving the overall performance of the ecosystem.

The last one is the creation of a new agent economic system, leveraging the combination of AI and blockchain technology to open up new economic models, allowing AI agents to play key roles in content creation, financial services, and more. A typical case is MyShellAI, which enables global players to train their own AI agents from game clips, thus creating a new on-chain Agent Economy. This indicates that a brand new economic system is forming, where AI agents can not only participate in content creation and game interaction but also act as independent economic entities, promoting the development of the creative economy and the integration of AI in the crypto field.

In terms of decentralized computing power, solutions such as secure multi-party computation, differential privacy, and federated learning are provided to meet the training and inference needs of large AI models. Facing technical challenges and bottlenecks, decentralized computing power has greater opportunities for implementation in inference. io.net has demonstrated how to reduce resource costs while ensuring economic benefits through token incentives and hardware aggregation, providing ML teams with ultra-scalable GPU resources. Decentralized computing networks will continue to explore and may provide effective solutions for large-scale AI training and inference.

Regarding data, OKX Ventures believes it mainly includes decentralized AI data markets, data incentive mechanisms, and privacy protection and data rights confirmation. In decentralized AI data markets, transparency, immutability, and traceability of data transactions are achieved, while smart contracts automatically execute transactions and allocate profits, enhancing market efficiency. Data incentive mechanisms promote the self-growth and prosperity of the ecosystem by tokenizing rewards for participants who contribute data and computing resources.

Privacy protection and data rights confirmation ensure the confidentiality and ownership of user data through cryptographic algorithms and zero-knowledge proofs, ensuring users have control over their data. These trends collectively drive the advancement and application of decentralized AI technology, forming a more secure, efficient, and fair digital economic environment. Here, 0xScope fine-tuned the Crypto industry model based on on-chain data and blockchain industry data, generating practical corresponding value.

In summary, the Crypto + AI sector has spawned numerous subfields, including infrastructure-level model training, data, decentralized computing power, and hardware, as well as model-level AI agents, ZKML, FHE, and application-level metaverse, gaming, AIGC, and memes, along with blockchain-level RAAS, oracles, co-processors, UBI, etc. Among these, projects at the infrastructure and application levels are developing rapidly, such as io.net in computing power, Flock in foundational models, ZeroGravity in blockchain infrastructure, Myshell in AI agents, and 0xScope in application layers.

Overview of AI Projects Invested by OKX Ventures

Currently, OKX Ventures has invested in numerous innovative projects in the Crypto + AI sector, providing diversified and comprehensive support not only in funding but also in technology, community, and resource connections.

0xscope

• Project Overview
• 0xScope is the AI Data layer for Web3, focusing on providing datasets suitable for Agent training and a knowledge base supporting Agent operations for Web3 projects. As of now, most mainstream Web3 AI DAPPs are using 0xScope's data services.
• Project Highlights
• 0xScope has launched data analysis products such as ScopeVC and ScopeScan. As of now, the total number of users has exceeded 500,000.
• Based on the current data foundation, it has launched the Web3 AI Q&A Agent—Scopechat, which is currently the only data analysis Agent on the market that can answer real-time knowledge data as well as ONCHAIN data. ScopeChat has already integrated with OKX Wallet, with a total user count exceeding 500,000 and DAU exceeding 10,000 (excluding witch users).
• In the next phase, 0xScope will collaborate with Ritual, Flashbots, and Chainlink to create a universal decentralized Web3 AI Agent, addressing the accessibility issues in Web3.
• Overall, 0xScope's development has gone through five stages of upgrades: data analysis tools, data infrastructure, AI data analysis, AI data infrastructure, and universal Agents, continuously deepening its moat around high-quality data foundations. The team is actively exploring the market based on user needs and technological evolution, while also receiving positive market feedback.

Flock

• Project Overview
• FLock is an innovative, end-to-end AI co-creation platform that integrates decentralized machine learning technology onto the blockchain. The platform redefines the processes of AI model training, fine-tuning, and inference, enabling the public to contribute knowledge and enrich AI models while protecting privacy. FLock's vision is to democratize AI technology through community-owned models, privacy-protecting data, and decentralized computing as its three foundational pillars. The platform encourages the creation of diverse AI models, from intelligent agents to complex trading and aggregation robots, ushering in a new era of accessible AI tools.
• Project Highlights
• Features include community-owned models, privacy-protecting data, and decentralized computing. FLock employs cutting-edge technologies such as federated learning, zkFL, homomorphic encryption, and secure multi-party computation (SMPC) to protect user data privacy and train and fine-tune AI models through decentralized computing power. Additionally, FLock encourages community participation in the expansion and refinement of AI models through its co-creation platform and provides a federated learning client to developers, supporting decentralized model training and validation while adhering to its commitment to data privacy.
• FLock addresses the issues of limited access, model bias, and unfair compensation for data contributors caused by centralized control of AI models. Through decentralized learning, local models are trained on decentralized training nodes and synchronized with global models, ensuring user data remains securely stored locally. FLock also employs technologies such as zkFL, homomorphic encryption, and secure multi-party computation to provide additional protection for data privacy and implements governance over the entire training process via blockchain, which is automatic, transparent, and immutable.
• Through these features, FLock achieves decentralized machine learning, solving a series of problems caused by centralized control of AI models.

Zero Gravity

• Project Overview
• ZeroGravity (abbreviated as 0G) is the first data availability system with a built-in universal storage layer, featuring strong scalability and decentralization. It achieves scalability by dividing the data availability workflow into data publishing channels and data storage channels. A large amount of data transmission occurs in the data storage channels supported by the storage layer, which achieves horizontal scaling through carefully designed partitioning, while the data publishing channels ensure data availability properties through consensus based on data availability sampling, requiring only a minimal amount of data to flow through the consensus protocol, thus avoiding broadcast bottlenecks.
• ZeroGravity's current product design includes the Data Availability Layer (0G DA) system and the Storage Layer (0G Storage).
• Project Highlights
• Features of 0G DA (Data Availability): The 0G DA system is a scalable data availability (DA) service layer built directly on top of decentralized storage systems, solving scalability issues by minimizing the amount of data transmission required for broadcasting. Its universal decentralized storage design also enables it to support various types of availability data from diverse scenarios, not limited to second-layer networks but also including decentralized AI infrastructure. 0G DA focuses on providing high security and high throughput data availability guarantees.
• Features of 0G Storage (Storage Layer): The storage layer of 0G consists of a storage network connected to a separate consensus network. Each storage node actively participates in the mining process by submitting proofs of data accessibility for specific data to smart contracts deployed on the consensus network. Once the smart contract verifies the proof, the storage node is rewarded accordingly. By providing more rewards to nodes for specific data within the same partition as the storage, partitioning is enabled. This incentive-based mechanism rewards nodes for their contributions rather than punishing them for misconduct, thereby better encouraging node participation in the maintenance of the network, which can practically promote better scalability. 0G storage is also designed as a universal storage system with multiple layers of abstraction and structure, including an append-only log layer for archiving unstructured data and a key-value layer for managing variable and structured data. This allows 0G to support reliable data indexing and a wider variety of availability data types from Layer 2 and AI scenarios.

Myshell

• Project Overview
• MyShell was born out of a profound reflection on the closed-source model ecosystem that has emerged in the current development of artificial intelligence technology. MyShell's vision is to create a fair and open AI ecosystem, a place where every participant can find their value. MyShell advocates for the democratization of AI, including supporting the development of an AI industry centered around open-source models, stimulating the vitality and innovative potential of the entire ecosystem through comprehensive incentive mechanisms, ensuring that creators and users receive their rightful benefits for their data and creations, and allowing users and developers to freely choose different AI models without worrying about dependence on a single model and monopoly, including the choice of decentralized computing and storage.
• Project Highlights
• Open-source model support and an open-compatible development platform: MyShell focuses on promoting the prosperity of AI technology through multi-contributor efforts, facilitating the development of the AI industry centered around open-source models. Creators can use the same method to call different models and external APIs, while third-party model and API providers can integrate their services into the platform for developers to use, lowering the barriers to innovation.
• Ecological incentive mechanisms and value redistribution: Ensuring that creators and users receive their rightful benefits for their data and creations through a transparent benefit distribution mechanism that redistributes the value generated within the ecosystem; a transparent reward system that fairly allocates rewards to contributors within the ecosystem, including model providers, application developers, users, investors, etc. Through comprehensive incentive mechanisms, the entire ecosystem's vitality and innovative potential are stimulated, maximizing the passion for innovation and contribution.
• Application creation and distribution promotion platform: An app store that helps creators publish and manage AI applications, where creators can see trends in popular AI applications and purchase advertisements to promote their works.
• Freedom of choice: Allowing users and developers to freely choose different AI models and decentralized computing and storage, avoiding dependence on any single model or technology and potential monopolies.
• Currently, Myshell's GitHub has 9k Stars and 1.3k Forks, with daily active peaks exceeding 100k, making it the most active area for creators and users in the AI + Crypto field.

io.net

• Project Overview
• io.net is a pioneering decentralized computing network based on Solana, designed to allow machine learning engineers to access distributed cloud cluster resources at a cost far lower than traditional centralized services. By aggregating underutilized GPU resources from independent data centers, crypto miners, and crypto projects like Filecoin and Render, io.net creates a decentralized physical infrastructure network (DePIN), providing engineers with substantial computing power that is both efficient and easy to implement, meeting a wide range of task needs from preprocessing, distributed training to hyperparameter tuning, reinforcement learning, and model serving.
• Project Highlights
• io.net is a pioneering decentralized computing network based on Solana, aimed at allowing machine learning engineers to access distributed cloud clusters at extremely low costs. The platform's mission is to integrate one million GPUs within a decentralized physical infrastructure network (DePIN), providing engineers with large-scale computing power that surpasses traditional centralized services in cost-effectiveness, customizability, and ease of implementation.
• Cost efficiency: io.net can provide computing costs up to 90% lower than traditional cloud service providers like AWS by utilizing underutilized resources such as GPUs from independent data centers and crypto miners.
• Quick access and deployment: Compared to traditional cloud providers, io.net users can access and deploy clusters within 90 seconds without facing cumbersome KYC processes or long-term contracts.
• Flexibility and control: io.net offers high flexibility and user control, supporting a wide range of GPU and CPU options, including NVIDIA RTX series, AMD Ryzen series, and Apple's M2 chip with unparalleled neural engines.
• Optimized for machine learning: io.net is built on the ray.io machine learning framework, supporting a wide range of machine learning workloads, from reinforcement learning to deep learning, tuning, and model serving, meeting the needs of machine learning engineers.
• Decentralized physical infrastructure network (DePIN): As the first and only GPU DePIN, io.net provides a platform connecting computing power providers and users for machine learning and all GPU use cases. io.net solves this issue by aggregating underutilized GPU resources from independent data centers, crypto miners, and crypto projects like Filecoin and Render. These resources are combined in the decentralized physical infrastructure network (DePIN), enabling engineers to access substantial computing power.
• io.net supports various tasks such as batch inference, model serving, parallel training, parallel hyperparameter tuning, and reinforcement learning, aiming to provide general computing services for Python workloads. Since the Solana Hackathon in February 2023 and the Solana Austin Hacker House, io.net has demonstrated its powerful capabilities and efficiency in providing distributed GPU networks for machine learning teams in training and inference workflows.

Diversified Empowerment, OKX Ventures Will Continue to Support Crypto + AI Innovative Projects

Crypto + AI will play an important role in the development of the digital future, endowing various services and products with higher levels of intelligence and promoting the digitalization and intelligence processes across various fields. For example, in intelligent investment, risk management, and anti-fraud systems, it will enhance the intelligence level of financial services, reduce risks, and improve efficiency, bringing more convenience and opportunities to human society.

Technological innovations in the AI field are evolving rapidly, with new algorithms, models, and application scenarios continuously emerging, unleashing enormous market potential across various industries such as healthcare, finance, retail, and manufacturing. In particular, the integration and innovation of Crypto + AI bring more efficient, secure, and intelligent transactions and services to the crypto industry, driving the development and application of cryptocurrency technology while also providing new scenarios and opportunities for the application of AI technology.

As a frontline participant in the industry, OKX Ventures will continue to focus on the innovative track of the combination of AI and Crypto, adopting both pragmatic and forward-looking strategies, actively participating in market construction, and supporting quality innovative projects and entrepreneurs.

Although the current AI token market does not rule out the possibility of bubbles, OKX Ventures' long-term goal is to transcend this phase, deeply explore and invest in projects that can bring genuine technological innovation, solve practical problems in the AI field, and ultimately sustain survival and development in the market. We firmly believe that only those AI + Crypto projects that are technically groundbreaking and can provide substantial value will achieve true success and sustainable development in the future.